A computer network is a collection of interconnected computing devices that can exchange data and share resources. Example network devices include layer two devices that operate within the second layer (L2) of the Open Systems Interconnection (OSI) reference model, i.e., the data link layer, and layer three devices that operate within the third layer (L3) of the OSI reference model, i.e., the network layer. Network devices within computer networks often include a control unit that provides control plane functionality for the network device and forwarding components for routing or switching data units.
An Ethernet Virtual Private Network (EVPN) may be used to extend two or more remote layer two (L2) customer networks through an intermediate layer three (L3) network (usually referred to as a provider network), in a transparent manner, i.e., as if the intermediate L3 network does not exist. In particular, the EVPN transports L2 communications, such as Ethernet packets or “frames,” between customer networks via the intermediate network. In a typical configuration, provider edge (PE) network devices (e.g., routers and/or switches) coupled to the customer edge (CE) network devices of the customer networks define label switched paths (LSPs) within the provider network to carry encapsulated L2 communications as if these customer networks were directly attached to the same local area network (LAN). In some configurations, the PE network devices may also be connected by an IP infrastructure in which case IP/GRE tunneling or other IP tunneling can be used between the network devices.
As a PE network device in an EVPN forwards Ethernet frames, the PE network devices learn L2 state information for the L2 customer networks. The L2 state information may include media access control (MAC) addressing information for the CE network devices and customer equipment within the customer network and the physical ports of the PE network device through which the customer devices are reachable. The PE network devices typically store the MAC addressing information in L2 learning tables associated with each of their physical interfaces. When switching an individual Ethernet frame having a given destination MAC address, a PE network device typically broadcasts the Ethernet frame to all of its physical ports unless the PE network device has previously learned the specific physical port through which to the destination MAC address is reachable. In this case, the PE network device forwards a single copy of the Ethernet frame out the associated physical port.
In an EVPN, MAC learning between PE network devices occurs in the control plane rather than in the data plane (as happens with traditional bridging) using a routing protocol. For example, in EVPNs, a PE network device typically uses the Border Gateway Protocol (BGP) (i.e., an L3 routing protocol) to advertise to other provider edge network devices the MAC addresses learned from the local consumer edge network devices to which the PE network device is connected. A PE device may use BGP route advertisement message to announce reachability information for the EVPN, where the BGP route advertisement specifies one or more MAC addresses learned by the PE network device instead of L3 routing information.
In some configurations, when a link failure or other topology event occurs within an EVPN, the PE network device may send BGP withdrawal messages to other PE network devices to indicate that the PE network device is no longer able to forward network traffic for the particular MAC addresses previously associated with the link that has now failed. Consequently, in the case of network topologies in which a PE router has learned thousands or potentially millions of MAC addresses for a customer network, failure of a network link may cause the PE network device to send a significant number of withdrawal messages to neighboring PE network devices so as to withdraw each individual MAC address. This may, in turn, result in a high volume of control plane messages traversing the intermediate network of the EVPN and yield suboptimal convergence times after the link failure or other topology change.